Disclaimer: IANAL, IANADD, I haven't been actively engaged with debian-legal for very long, and my interpretation of the meaning of "derivative work" and its consequences for the scope of the GPL appears to contrast rather strongly with the FSF's and with some other debian-legal participants'. But I can point you to some US case law, and statutes elsewhere, which I believe to be relevant. Perhaps this will be helpful to you in assessing the quality of the local conventional wisdom.
On Fri, 28 Jan 2005 21:49:08 +0100, Frank Küster <[EMAIL PROTECTED]> wrote: [snip] > Let's assume a piece of technical documentation (standalone, i.e not > part of a software package; something like selfhtml or LaTeX's lshort), > is licensed under GPL, with an additional text stating what the > preferred form for modification is (say, LaTeX or docbook). > > As I understand it, anybody can take the text and publish it as a > printed book, as long as he also prints where the source code can be > found (GPL clause 3.b or 3.c). If he creates a derived work - for > example by extending each chapter, but keeping the structure and most of > the original text - he has to license that under GPL, too (and thus > provide the source code). That's correct. Publishing this book, or in general any action that would normally require a copyright license and isn't accompanied by a strong claim of alternate licensing terms, is fairly strong evidence of "acceptance through conduct" of the offer of contract contained in the GPL, in any jurisdiction where a non-exclusive copyright license can be created by this form of acceptance. That would include any jurisdiction where any retail software license is enforceable. There is a non-zero threshold for evidence of acceptance, however. See Specht v. Netscape, which held that a mere "browse-wrap" license on downloadable software was similar to an open box containing copies of a free newspaper, and that merely taking and using that newspaper doesn't prove acceptance of the attached license. Picking up a copy of a free newspaper doesn't convey license to republish its contents, however; so publishing printed copies of a GPL document is clearly well over this threshold. Judging from the two cases I have been able to find in which the GPL has come up in US court (namely, Progress Software v. MySQL and Planetary Motion v. Techplosion), there is no reason to think that there is any defect in the GPL as an enforceable offer of contract regarding material on which the copyright holder retains its ownership. Specifically, publication of a work with the GPL attached is not release into the public domain and does consitute "use in commerce", according to an argument in Planetary Motion (a valid appellate precedent) which appears to me also to prove sufficient consideration on both sides to form a bilateral contract. Notwithstanding the public claims of the FSF's general counsel, there is no such thing as a "non-contract license" in any jurisdiction I have heard named. With help from other debian-legal participants, I've verified this assertion against statute and/or precedent in at least the US, Canada, UK, Belgium, and Germany. I have also asked [EMAIL PROTECTED] about their stance, and received nothing but pointers to an interview with the general counsel and the 1709 Statute of Anne. The FSF's general counsel submitted an expert witness affidavit in the MySQL case, but it makes no reference to relevant modern legal precedent. In general, this means that a claim of copyright infringement cannot succeed unless the copyright holder can demonstrate, in addition to the usual criteria for infringement, that the defendant either exceeded the scope of the license granted in the contract or breached the contract to an extent that justifies rescission of the license. This is consistent with the judge's ruling in Progress Software vs. MySQL, in which MySQL did not succeed in obtaining a preliminary injunction on copyright grounds. > I have two questions > > 1. The first is whether there are any established criteria by which the > creation of a derived work can be distinguished from mere aggregation. > > I assume that if a book on the technical aspects of computer > typesetting would include, as an appendix, a GPL'ed text on > typography, this would be only aggregation. At least if typographical > questions don't play a role in the rest of the text. > > But what if there are extensive references to specific parts of the > appendix in the text? What if it is a chapter in that book? I'm going to try to answer the question I think you are asking: "Joe-Bob writes a book about GPL Work X, on which the copyright is held by Sue. He includes Work X as Chapter X (or Appendix X) of his book, and makes extensive references to Work X in Chapter Y (and possibly every other chapter). Can Sue successfully sue Joe-Bob to either block distribution of the book or to require release of the 'source code' for the entire book under the GPL?" I believe the answer to be No. For reasons that I discuss below, I don't think that the language of the GPL, taken together with applicable law and precedent, should be read to cover "collections" or "compilations", irrespective of how the undefined term "mere aggregation" is construed. In a software context, that means that I disagree with the FSF's claim that linking creates a derivative work. In a book context, I don't believe that accepting license under the GPL to copy Work X as a chapter or appendix forces application of the GPL to the entire book. As for the "extensive references", you can't claim license under the GPL to include bits of Work X in an integral work of authorship that is Chapter Y (or Chapters 1 through N) of your book, unless you are willing to consider Chapter Y a derivative work and to release it under the GPL. However, if Chapters 1 through N are a separate work of authorship (about which more later), combined with Work X to form a compilation, they may well be defensible under a theory entirely unrelated to the GPL, that of "fair use". I will come back to "fair use" below, focusing first on whether the GPL ought to be read to cover the whole book anyway. The GPL explicitly states in Section 0 that "a 'work based on the Program' means either the Program or any derivative work under copyright law". It provides a paraphrase of that definition, which includes the phrase "work containing the Program", but to read this automatically to mean the biggest possible "work" (i. e., the whole book, or the whole boxed set of which it is one volume) is contrary to its meaning in copyright law. The term "derivative work" is defined in the Berne Convention, to which the great majority of nations are signatories. It appears in that document once, as the title of Article 2, Clause 3, which says: "Translations, adaptations, arrangements of music and other alterations of a literary or artistic work shall be protected as original works without prejudice to the copyright in the original work." Articles 8 and 12 reserve to the author (or copyright holder) the right to authorize, respectively, the creation of translations and of the other listed forms of derivative work. "Collections of literary or artistic works such as encyclopaedias and anthologies" are discussed separately in Article 2, Clause 5, and are considered a separate category of "intellectual creations". It appears clear to me that they are not intended to be included in the category of "derivative works", but rather as a special case in which the only protectable element belonging to the maker of the collection is their "selection and arrangement". Unlike all of the listed forms of derivative work, collections are not mentioned in articles 8 and 12; the only other use of the word "collection" in the text of the convention is in Article 2bis, Clause 3, which states essentially that individual works placed into the public domain by their political or legal nature can still be selected and arranged to form a copyrightable anthology. Any given country's implementation of the Berne Convention may vary somewhat, but the US statute (at least as of 1986) and the case law I have seen are consistent with the interpretation that "compilations" (or the subset "collective works") are a disjoint category from "derivative works". See 17 USC 101, and compare the UK CDPA sections 3 (1) (a) (a "table" or "compilation" is a subtype of literary work) and 21 ("adaptations" are defined to cover essentially the same scope as Berne Convention "derivative works"). OK, now on to the text of the GPL. The "mere aggregation" clause that ends Section 2 of the GPL purports to be an exception to the requirement that larger works be entirely GPL. In fact, the preceding paragraph contains the only mention of "collective works" in the GPL, stating that "the intent is to exercise the right to control the distribution of derivative or collective works based on the Program". The paragraph before that, however, is worth quoting in its entirety: "These requirements apply to the modified work as a whole. If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when you distribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it." Remember that Section 0 defined a "work based on the Program" to be a derivative work, no more and no less. Section 2 also begins with language addressing modification to form a "work based on the Program". The above paragraph again talks about "a whole which is a work based on the Program". It's a close call, but I think it is correct to read the entirety of Section 2 as being applicable only to derivative works, and hence to permit Joe-Bob to accept license under the GPL to copy Work X into his compilation, together with another work that itself is legitimate under a "fair use" defense. It's important to note that what we are construing here is not a statute, it is an offer of contract. A US court is obliged, as a matter of law, to construe a statute in order to give each of its provisions force; if this rule were applicable, it might be necessary to rule that "work based on the Program" was meant to include collective works all along, since otherwise the "mere aggregation" clause would have little or no effect. A contract, on the other hand, must as a matter of law be construed against the offeror, and that goes double for a "standard-form" contract in which there is no "meeting of the minds" (back-and-forth negotiation) between offeror and offeree. The GPL invites the reader to interpret "derivative work" using its meaning under copyright law, and thus to accept license to modify and copy Work X, maintaining its status as a separately identifiable work of authorship, as part of a compilation which also contains a critical work for which no license is required. As I said, this is a close call, and the need for the "fair use" defense to justify the rest of Joe-Bob's book makes it closer in practice. Note also that the language of Section 2b, if read by itself, does not use the phrase "work based on the Program" but rather the more inclusive language "any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof". So there's some room for Sue to argue that Section 2b covers collective works even if nothing else in the GPL does; I think that this argument should fail, but a court might not. In the absence of the "mere aggregation" clause, the fact that Sue has winked at the creation of other collective works containing Work X, themselves not entirely covered by the GPL, could perhaps be used as a wholly separate argument to estop her from pursuing this particular use. But in the presence of that clause, if Joe-Bob's argument that the whole section only applies to derivative works were to fail, and if the court held that Sue could reasonably have considered other uses to have been "mere aggregation" while objecting to this use, then Joe-Bob's conduct could be ruled to have breached the contract formed under the GPL. If Joe-Bob cannot rely on valid license under the GPL to include Work X as Chapter (or Appendix) X of his book, then he has to fall back on the argument that inclusion of the entire text of Work X is still "fair use" without need for license. This is now a pure copyright infringement case and the GPL is out of the picture except as evidence of the limited commercial value of non-GPL licenses to Work X as a standalone work. Joe-Bob might still win, since the most important "fair use" factor is in his favor; the only effect his use could be said to have had on the market for Work X is to reduce Sue's ability to charge for authorizing a similar book by another author, and that author could write and publish that book anyway, printing Work X as a separate volume. But Joe-Bob's publisher might be more prudent to create this separation in the first place. If GPL section 2b is to be read broadly, the borderline of "mere aggregation" is anybody's guess. Is it OK to put a Work X pamphlet in a sleeve inside the book's back cover? Is it OK to have typographical uniformity but to ship the pamphlet to retailers under separate cover, or is that some kind of contributory infringement because the retailer or end user creates a "work" by combining them? You wind up with the same kind of hair-splitting that has occurred in the Eclipse+Kaffe debate. > 2. I fail to find the right technical or juridical terms here, but I > guess in most jurisdictions it is allowed to cite other texts, or to > publish a book that discusses some text in detail (like > interpretation of a poem, or detailed rebuttal of a scientific > paper). In such a case, the book would not exist without prior > existence of the original text. Would such a thing be regarded a > derived work, and would therefore a text published under GPL impose > restrictions that would not hold for a text published without a > license, simply in printed form? As I mentioned above, the applicable defense here is "fair use", and does not rely on any license terms but solely on statutory limits on the copyright monopoly. In the US at least, "fair use" operates as an affirmative defense after it is established that substantive "copying" of copyrightable fragments in excess of "de minimis" has taken place. The scope of "fair use" varies to some degree from jurisdiction to jurisdiction, but is generally held to permit, for instance, critical works that include fragments of an original work that are genuinely necessary for coherent commentary. Note that "fair use" is different from the argument that I have been making elsewhere for the application-that-uses-a-GPL-library case, which uses different precedents to arrive at the conclusion that the bits of header file or equivalent published interface that wind up in the application are uncopyrightable. These precedents address the functional aspect of the interface between a software component and its user, the hardware/firmware on which it runs, or another software component, and are only weakly applicable if at all to the case of critical commentary on a text. In the US, the criteria for "fair use" are defined in 17 USC 107 and well illustrated by appellate precedents. The criterion usually given the most weight is "the effect of the use upon the potential market for or value of the copyrighted work." A critical essay is rarely held to detract from the market for the original work, and is only likely to fail as "fair use" if it incorporates the bulk of the original without license. Critical essays are routinely appended to original works (published under license from the copyright holder) to form critical editions without requiring additional license beyond that for publication of the original. So if 1) Joe-Bob has Sue's license under the GPL to distribute Work X, 2) the rest of Joe-Bob's book is defensible as "fair use" without any need for license, and 3) it is correct to interpret the GPL to cover Work X while not reaching out to grab the entire book (which may depend on the details of bookbinding and distribution), then Joe-Bob can successfully defend against Sue's claim of copyright infringement and/or breach of contract. Whew. IANAL, etc., etc. Cheers, - Michael